Pyrolysis patterns of 5 close Corynebacterium species analyzed by artificial neural networks

In the present study, an artificial neural network was trained with the Stuttgart Neural Networks Simulator, in order to identify Corynebacterium species by analyzing their pyrolysis patterns. An earlier study described the combination of pyrolysis, gas chromatography and atomic emission detection we used on whole cell bacteria. Carbon, sulfur and nitrogen were detected in the pyrolysis compounds. Pyrolysis patterns were obtained from 52 Corynebacterium strains belonging to 5 close species. These data were previously analyzed by Euclidean distances calculation followed by Unweighted Pair Group Method of Averages, a clustering method. With this early method, strains from 3 of the 5 species (C. xerosis, C. freneyi and C. amycolatum) were correctly characterized even if the 29 strains of C. amycolatum were grouped into 2 subgroups. Strains from the 2 remaining species (C. minutissimum and C. striatum) cannot be separated. To build an artificial neural network, able to discriminate the 5 previous species, the pyrolysis data of 42 selected strains were used as learning set and the 10 remaining strains as testing set. The chosen learning algorithm was Back-Propagation with Momentum. Parameters used to train a correct network are described here, and the results analyzed. The obtained artificial neural network has the following cone-shaped structure: 144 nodes in input, 25 and 9 nodes in 2 successive hidden layers, and then 5 outputs. It could classify all the strains in their species group. This network completes a chemotaxonomic method for Corynebacterium identification.     

Rapid identification of streptomycetes by artificial neural network analysis of pyrolysis mass spectra

Abstract An artificial neural network was trained to distinguish between three putatively novel species of Streptomyces using normalised, scaled pyrolysis mass spectra from three representative strains of each of the taxa, each sampled in triplicate. Once trained, the artificial neural network was challenged with spectral data from the original organisms, the 'training set', from additional members of the putative novel taxa and from over a hundred strains representing six other actinomycete genera. All of the streptomycetes were correctly identified but many of the other actinomycetes were mis-identified. A modified network topology was developed to recognise the mass spectral patterns of the non-streptomycete strains. The resultant neural network correctly identified the streptomycetes, whereas all of the remaining actinomycetes were recognised as unknown organisms. The improved artificial neural network provides a rapid, reliable and cost-effective method of identifying members of the three target streptomycete taxa.     

Rapid identification using pyrolysis mass spectrometry and artificial neural networks of Propionibacterium acnes isolated from dogs

Curie-point pyrolysis mass spectra were obtained from reference Propionibacterium strains and canine isolates. Artificial neural networks (ANNs) were trained by supervised learning (with the back-propagation algorithm) to recognize these strains from their pyrolysis mass spectra; all the strains isolated from dogs were identified as human wild type P. acnes. This is an important nosological discovery, and demonstrates that the combination of pyrolysis mass spectrometry and ANNs provides an objective, rapid and accurate identification technique. Bacteria isolated from different biopsy specimens from the same dog were found to be separate strains of P. acnes , demonstrating a within-animal variation in microflora. The classification of the canine isolates by Kohonen artificial neural networks (KANNs) was compared with the classical multivariate techniques of canonical variates analysis and hierarchical cluster analysis, and found to give similar results. This is the first demonstration, within microbiology, of KANNs as an unsupervised clustering technique which has the potential to group pyrolysis mass spectra both automatically and relatively objectively.     

Differentiation of probiotic and environmental Saccharomyces cerevisiae strains in animal feed

Aims: To establish an identification system for probiotic Saccharomyces cerevisiae strains based on artificial neural network (ANN)–assisted Fourier‐transform infrared (FTIR) spectroscopy to improve quality control of animal feed. Methods and Results: The ANN‐based system for differentiating environmental from probiotic S. cerevisiae strains comprises five authorized feed additive strains plus environmental strains isolated from different habitats. A total of 108 isolates were used as reference strains to create the ANN. DHPLC analysis and δ‐PCR were used as reference methods to type probiotic yeast isolates. The performance of the FTIR‐ANN was tested in an internal validation using unknown spectra of each reference strain. This validation step yielded a classification rate of 99·1 %. For an external validation, a test data set comprising 965 spectra of 63 probiotic and environmental S. cerevisiae isolates unknown to the ANN was used, resulting in a classification rate of 98·2 %. Conclusions: Our results demonstrate that probiotic S. cerevisiae strains in feed can be differentiated successfully from environmental isolates using both genotypic approaches and ANN‐based FTIR spectroscopy. Significance and Impact of the Study: FTIR‐based artificial neural network analysis provides a rapid and inexpensive technique for yeast identification both at the species and at the strain level in routine diagnostic laboratories, using a single sample preparation.     

Rapid authentication of animal cell lines using pyrolysis mass spectrometry and auto-associative artificial neural networks

Summary Pyrolysis mass spectrometry (PyMS) was used to produce biochemical fingerprints from replicate frozen cell cultures of mouse macrophage hybridoma 2C11-12, human leukaemia K562, baby hamster kidney BHK 21/C13, and mouse tumour BW-O, and a fresh culture of Chinese hamster ovary CHO cells. The dimensionality of these data was reduced by the unsupervised feature extraction pattern recognition technique of auto-associative neural networks. The clusters observed were compared with the groups obtained from the more conventional statistical approaches of hierarchical cluster analysis. It was observed that frozen and fresh cell line cultures gave very different pyrolysis mass spectra. When only the frozen animal cells were analysed by PyMS, auto-associative artificial neural networks (ANNs) were employed to discriminate between them successfully. Furthermore, very similar classifications were observed when the same spectral data were analysed using hierarchical cluster analysis. We demonstrate that this approach can detect the contamination of cell lines with low numbers of bacteria and fungi; this approach could plausibly be extended for the rapid detection of mycoplasma infection in animal cell lines. The major advantages that PyMS offers over more conventional methods used to type cell lines and to screen for microbial infection, such as DNA fingerprinting, are its speed, sensitivity and the ability to analyse hundreds of samples per day. We conclude that the combination of PyMS and ANNs can provide a rapid and accurate discriminatory technique for the authentication of animal cell line cultures.     

Rapid quantitative analysis of binary mixtures of Escherichia coli strains using pyrolysis mass spectrometry with multivariate calibration and artificial neural networks

Pyrolysis mass spectrometry (PyMS) and multivariate calibration were used to show the high degree of relatedness between Escherichia coli HB101 and E. coli UB5201. Next, binary mixtures of these two phenotypically closely related E. coli strains were prepared and subjected to PyMS. Fully interconnected feedforward artificial neural networks (ANNs) were used to analyse the pyrolysis mass spectra to obtain quantitative information representative of the level of E. coli UB5201 in E. coli HB101. The ANNs exploited were trained using the standard back propagation algorithm, and the nodes used sigmoidal squashing functions. Accurate quantitative information was obtained for mixtures with >3% E. coli UB5201 in E. coli HB101. To remove noise from the pyrolysis mass spectra and so lower the limit of detection, the spectra were reduced using principal components analysis (PCA) and the first 13 principal components used to train ANNs. These PCA-ANNs allowed accurate estimates at levels as low as 1% E. coli UB5201 in E. coli HB101 to be predicted. In terms of bacterial numbers, it was shown that the limit of detection for PyMS in conjunction with ANNs was 3 × 10⁴ E. coli UB5201 cells in 1·6 × 10⁷ E. coli HB101 cells. It may be concluded that PyMS with ANNs provides a powerful and rapid method for the quantification of mixtures of closely related bacterial strains.     

Resolution of batch variations in pyrolysis mass spectrometry of bacteria by the use of artificial neural network analysis

A simple, but stringent, three group model of bacterial interstrain identity (two cultures of the same strain ofEscherichia coli) and difference (a culture of a serologically distinct strain) was used in multiple serial weekly subcultures for five weeks to demonstrate the effect of both growth-related (phenotypic) and machine-related variation on pyrolysis mass spectra. An aliquot of serum from a single sample was included in each pyrolysis batch to distinguish machine drift from culture drift. Conventional principal component (PC) canonical variate (CV) analysis was successful within each pyrolysis batch but the variations between batches precluded the use of data from more than one batch in successful PCCV analysis. In contrast, artificial neural networks (ANNs) trained with data from one batch could be successfully used to identify groups in data from non-contemporaneous pyrolysis batches. Although the ANN method will require validation in more complex settings than this simple model, it is a promising approach to the problem of batch constraint in pyrolysis mass spectrometry.     

Artificial intelligence based optimization of exocellular glucansucrase production from Leuconostoc dextranicum NRRL B-1146

Two different artificial intelligence techniques namely artificial neural network (ANN) and genetic algorithm (GA) were integrated for optimizing fermentation medium for the production of glucansucrase. The experimental data reported in a previous study were used to build the neural network. The ANN was trained using the back propagation algorithm. The ANN predicted values showed good agreement with the experimentally reported ones from a response surface based experiment. The concentrations of three medium components: viz Tween 80, sucrose and K(2)HPO(4) served as inputs to the neural network model and the enzyme activity as the output of the model. A model was generated with a coefficient of correlation (R(2)) of 1.0 for the training set and 0.90 for the test data. A genetic algorithm was used to optimize the input space of the neural network model to find the optimum settings for maximum enzyme activity. This artificial neural network supported genetic algorithm predicted a maximum glucansucrase activity of 6.92U/ml at medium composition of 0.54% (v/v) Tween 80, 5.98% (w/v) sucrose and 1.01% (w/v) K(2)HPO(4). ANN-GA predicted model gave a 6.0% increase of enzyme activity over the regression based prediction for optimized enzyme activity. The maximum enzyme activity experimentally obtained using the ANN-GA designed medium was 6.75+/-0.09U/ml which was in good agreement with the predicted value.     

Functional nodes in dynamic neural networks for bioprocess modelling

This contribution presents a novel method for the direct integration of a-priori knowledge in a neural network and its application for the online determination of a secondary metabolite during industrial yeast fermentation. Hereby, existing system knowledge is integrated in an artificial neural network (ANN) by means of 'functional nodes'. A generalized backpropagation algorithm is presented. For illustration, a set of ordinary differential equations describing the diacetyl formation and degradation during the cultivation is incorporated in a functional node and integrated in a dynamic feedforward neural network in a hybrid manner. The results show that a hybrid modelling approach exploiting available a-priori knowledge and experimental data can considerably outperform a pure data-based modelling approach with respect to robustness, generalization and necessary amount of training data. The number of training sets were decreased by 50%, obtaining the same accuracy as in a conventional approach. All incorrect decisions, according to defined cost criteria obtained with the conventional ANN, were avoided.     

Curie-point pyrolysis mass spectrometry applied to characterization and identification of selected Bacillus species

The use of pyrolysis mass spectrometry in the characterization and identification of Bacillus species was studied. Fifty-three strains of four closely related groups, Bacillus subtilis, B. pumilus, B. licheniformis and 'B. amyloliquefaciens', were used in a study of both sporulated and nonsporulated cultures. Pyrolysis was carried out using a Pyromass 8-80, a novel pyrolysis mass spectrometer specifically designed for fingerprinting complex samples. The pyrolysis data obtained were analysed using multivariate statistical techniques. All four groups could be differentiated using data from non-sporulated cultures but the data from sporulated cultures did not separate B. subtilis from 'B. amyloliquefaciens' or B. pumilus. In contrast, B. licheniformis was more clearly differentiated from the other three species using these data. Culture maturity affected the mass spectra obtained from non-sporulated cultures.     

Differentiation of Listeria monocytogenes serovars by using artificial neural network analysis of Fourier-transformed infrared spectra

A classification system based on Fourier transform infrared (FTIR) spectroscopy combined with artificial neural network analysis was designed to differentiate 12 serovars of Listeria monocytogenes using a reference database of 106 well-defined strains. External validation was performed using a test set of another 166 L. monocytogenes strains. The O antigens (serogroup) of 164 strains (98.8%) could be identified correctly, and H antigens were correctly determined in 152 (91.6%) of the test strains. Importantly, 40 out of 41 potentially epidemic serovar 4b strains were unambiguously identified. FTIR analysis is superior to PCR-based systems for serovar differentiation and has potential for the rapid, simultaneous identification of both species and serovar of an unknown Listeria isolate by simply measuring a whole-cell infrared spectrum.     

Pyrolysis mass spectrometry characterisation and numerical taxonomy ofAeromonas spp.

Reference strains (2) and 29 isolates ofAeromonas spp. from clinical material and environmental specimens were characterised in traditional biochemical tests, and in pyrolysis mass spectrometry, which gives data reflecting whole-cell composition. Numerical taxonomic analyses of the data sets were compared with conventional identification at species level, and pathogenic potential, as inferred from the origin of the isolates. Clustering with conventional test reaction patterns showed, for each of the species represented, a clearly defined core group of typical isolates, surrounded by a halo of aberrant strains. One further cluster comprised strains intermediate betweenA. caviae andA. hydrophila, and one strain was grossly atypical in both analyses. Clustering from pyrolysis data corresponded less well with species identification. Broadly, the biochemical division between core and halo strains was supported in pyrolysis forA. caviae andA. sobria, but the main group ofA. hydrophila in pyrolysis comprised strains clustering in the core and halo groups of this species, and three strains intermediate betweenA. hydrophila andA. caviae in biochemical tests. Two further pyrolysis clusters comprised core and halo strains ofA. hydrophila. However, pyrolysis clustering correlated well with inferred pathogenicity, showing four clusters of probable pathogens, six clusters of probable nonpathogens, and one two member cluster of doubtful status. Most strains that clustered in the species haloes, or in species-intermediate groups in biochemical tests, were non-human isolates, or were isolated in the absence of symptomatic infection. The correlation of inferred pathogenicity with biochemical clustering was poorer than that with pyrolysis clustering.Abbreviations CTRP conventional test reaction pattern - PyMS pyrolysis mass spectrometry     

Feedforward neural network models for handling class overlap and class imbalance

This paper proposes a framework for training feedforward neural network models capable of handling class overlap and imbalance by minimizing an error function that compensates for such imperfections of the training set. A special case of the proposed error function can be used for training variance-controlled neural networks (VCNNs), which are developed to handle class overlap by minimizing an error function involving the class-specific variance (CSV) computed at their outputs. Another special case of the proposed error function can be used for training class-balancing neural networks (CBNNs), which are developed to handle class imbalance by relying on class-specific correction (CSC). VCNNs and CBNNs are compared with conventional feedforward neural networks (FFNNs), quantum neural networks (QNNs), and resampling techniques. The properties of VCNNs and CBNNs are illustrated by experiments on artificial data. Various experiments involving real-world data reveal the advantages offered by VCNNs and CBNNs in the presence of class overlap and class imbalance.     

Principal component analysis and artificial neural network analysis of oral tissue fluorescence spectra: classification of normal premalignant and malignant pathological conditions

Pulsed laser-induced autofluorescence spectroscopic studies of pathologically certified normal, premalignant, and malignant oral tissues were carried out at 325 nm excitation. The spectral analysis and classification for discrimination among normal, premalignant, and malignant conditions were performed using principal component analysis (PCA) and artificial neural network (ANN) separately on the same set of spectral data. In case of PCA, spectral residuals, Mahalanobis distance, and scores of factors were used for discrimination among normal, premalignant, and malignant cases. In ANN, parameters like mean, spectral residual, standard deviation, and total energy were used to train the network. The ANN used in this study is a classical multiplayer feed-forward type with a back-propagation algorithm for the training of the network. The specificity and sensitivity were determined in both classification schemes. In the case of PCA, they are 100 and 92.9%, respectively, whereas for ANN they are 100 and 96.5% for the data set considered.